AI guidelines for mentoring programs (like GSoC and Outreachy)

AI guidelines for mentoring programs (like GSoC and Outreachy)

[Christian Couder]

Hi everyone,

We now have a new "AI guidelines" on
https://git.github.io/General-Application-Information/ following some
discussions between GSoC 2025 potential mentors and org admins by
email and on https://github.com/git/git.github.io/pull/771.

We came up relatively quickly with these guidelines because this year
71 out of 79 proposals we received for the GSoC 2025 were spammy and
very often AI generated. This is a significant increase compared to
previous years. I remember that a few years ago there were less than a
dozen spammy proposals. We also received some AI generated spam
patches and emails to our personal email addresses.

Other organizations participating in the GSoC 2025 also received an
increasing amount of such spam, and organizations which documented
some guidelines against it said they receive less of it. A large
amount of discussion has happened on the GSoC mentors list about this.

Let us know if you have an opinion about this, or if you think that
the whole Git project should have AI guidelines.

Thanks.

Re: AI guidelines for mentoring programs (like GSoC and Outreachy)

[Patrick Steinhardt]

Hi Chris,

On Tue, Apr 15, 2025 at 11:21:53AM +0200, Christian Couder wrote:
> Hi everyone,
> 
> We now have a new "AI guidelines" on
> https://git.github.io/General-Application-Information/ following some
> discussions between GSoC 2025 potential mentors and org admins by
> email and on https://github.com/git/git.github.io/pull/771.
> 
> We came up relatively quickly with these guidelines because this year
> 71 out of 79 proposals we received for the GSoC 2025 were spammy and
> very often AI generated. This is a significant increase compared to
> previous years. I remember that a few years ago there were less than a
> dozen spammy proposals. We also received some AI generated spam
> patches and emails to our personal email addresses.
> 
> Other organizations participating in the GSoC 2025 also received an
> increasing amount of such spam, and organizations which documented
> some guidelines against it said they receive less of it. A large
> amount of discussion has happened on the GSoC mentors list about this.

Yeah, it's been a bit of a pain this year indeed. Thanks for creating
the AI guidelines, let's hope it improves the situation.

> Let us know if you have an opinion about this, or if you think that
> the whole Git project should have AI guidelines.

We (you and I) have discussed this internally and got to the conclusion
that it's not needed for the Git project as a whole at the current point
in time. We haven't yet seen any obvious issues with AI-generated
patches on the mailing list, so it would probably be premature to worry
about it now already. We can and should reevaluate though in case we
ever see an uptick of slop.

I'm of course happy to hear differing opinions.

Patrick